1/*******************************************************************************
2
3  Intel PRO/10GbE Linux driver
4  Copyright(c) 1999 - 2006 Intel Corporation.
5
6  This program is free software; you can redistribute it and/or modify it
7  under the terms and conditions of the GNU General Public License,
8  version 2, as published by the Free Software Foundation.
9
10  This program is distributed in the hope it will be useful, but WITHOUT
11  ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  more details.
14
15  You should have received a copy of the GNU General Public License along with
16  this program; if not, write to the Free Software Foundation, Inc.,
17  51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
18
19  The full GNU General Public License is included in this distribution in
20  the file called "COPYING".
21
22  Contact Information:
23  Linux NICS <linux.nics@intel.com>
24  e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
25  Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
26
27*******************************************************************************/
28
29#include "ixgb.h"
30
31char ixgb_driver_name[] = "ixgb";
32static char ixgb_driver_string[] = "Intel(R) PRO/10GbE Network Driver";
33
34#ifndef CONFIG_IXGB_NAPI
35#define DRIVERNAPI
36#else
37#define DRIVERNAPI "-NAPI"
38#endif
39#define DRV_VERSION		"1.0.126-k2"DRIVERNAPI
40char ixgb_driver_version[] = DRV_VERSION;
41static char ixgb_copyright[] = "Copyright (c) 1999-2006 Intel Corporation.";
42
43/* ixgb_pci_tbl - PCI Device ID Table
44 *
45 * Wildcard entries (PCI_ANY_ID) should come last
46 * Last entry must be all 0s
47 *
48 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
49 *   Class, Class Mask, private data (not used) }
50 */
51static struct pci_device_id ixgb_pci_tbl[] = {
52	{INTEL_VENDOR_ID, IXGB_DEVICE_ID_82597EX,
53	 PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
54	{INTEL_VENDOR_ID, IXGB_DEVICE_ID_82597EX_CX4,
55	 PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
56	{INTEL_VENDOR_ID, IXGB_DEVICE_ID_82597EX_SR,
57	 PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
58	{INTEL_VENDOR_ID, IXGB_DEVICE_ID_82597EX_LR,
59	 PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
60
61	/* required last entry */
62	{0,}
63};
64
65MODULE_DEVICE_TABLE(pci, ixgb_pci_tbl);
66
67/* Local Function Prototypes */
68
69int ixgb_up(struct ixgb_adapter *adapter);
70void ixgb_down(struct ixgb_adapter *adapter, boolean_t kill_watchdog);
71void ixgb_reset(struct ixgb_adapter *adapter);
72int ixgb_setup_tx_resources(struct ixgb_adapter *adapter);
73int ixgb_setup_rx_resources(struct ixgb_adapter *adapter);
74void ixgb_free_tx_resources(struct ixgb_adapter *adapter);
75void ixgb_free_rx_resources(struct ixgb_adapter *adapter);
76void ixgb_update_stats(struct ixgb_adapter *adapter);
77
78static int ixgb_init_module(void);
79static void ixgb_exit_module(void);
80static int ixgb_probe(struct pci_dev *pdev, const struct pci_device_id *ent);
81static void __devexit ixgb_remove(struct pci_dev *pdev);
82static int ixgb_sw_init(struct ixgb_adapter *adapter);
83static int ixgb_open(struct net_device *netdev);
84static int ixgb_close(struct net_device *netdev);
85static void ixgb_configure_tx(struct ixgb_adapter *adapter);
86static void ixgb_configure_rx(struct ixgb_adapter *adapter);
87static void ixgb_setup_rctl(struct ixgb_adapter *adapter);
88static void ixgb_clean_tx_ring(struct ixgb_adapter *adapter);
89static void ixgb_clean_rx_ring(struct ixgb_adapter *adapter);
90static void ixgb_set_multi(struct net_device *netdev);
91static void ixgb_watchdog(unsigned long data);
92static int ixgb_xmit_frame(struct sk_buff *skb, struct net_device *netdev);
93static struct net_device_stats *ixgb_get_stats(struct net_device *netdev);
94static int ixgb_change_mtu(struct net_device *netdev, int new_mtu);
95static int ixgb_set_mac(struct net_device *netdev, void *p);
96static irqreturn_t ixgb_intr(int irq, void *data);
97static boolean_t ixgb_clean_tx_irq(struct ixgb_adapter *adapter);
98
99#ifdef CONFIG_IXGB_NAPI
100static int ixgb_clean(struct net_device *netdev, int *budget);
101static boolean_t ixgb_clean_rx_irq(struct ixgb_adapter *adapter,
102				   int *work_done, int work_to_do);
103#else
104static boolean_t ixgb_clean_rx_irq(struct ixgb_adapter *adapter);
105#endif
106static void ixgb_alloc_rx_buffers(struct ixgb_adapter *adapter);
107void ixgb_set_ethtool_ops(struct net_device *netdev);
108static void ixgb_tx_timeout(struct net_device *dev);
109static void ixgb_tx_timeout_task(struct work_struct *work);
110static void ixgb_vlan_rx_register(struct net_device *netdev,
111				  struct vlan_group *grp);
112static void ixgb_vlan_rx_add_vid(struct net_device *netdev, uint16_t vid);
113static void ixgb_vlan_rx_kill_vid(struct net_device *netdev, uint16_t vid);
114static void ixgb_restore_vlan(struct ixgb_adapter *adapter);
115
116#ifdef CONFIG_NET_POLL_CONTROLLER
117/* for netdump / net console */
118static void ixgb_netpoll(struct net_device *dev);
119#endif
120
121static pci_ers_result_t ixgb_io_error_detected (struct pci_dev *pdev,
122	                     enum pci_channel_state state);
123static pci_ers_result_t ixgb_io_slot_reset (struct pci_dev *pdev);
124static void ixgb_io_resume (struct pci_dev *pdev);
125
126/* Exported from other modules */
127extern void ixgb_check_options(struct ixgb_adapter *adapter);
128
129static struct pci_error_handlers ixgb_err_handler = {
130	.error_detected = ixgb_io_error_detected,
131	.slot_reset = ixgb_io_slot_reset,
132	.resume = ixgb_io_resume,
133};
134
135static struct pci_driver ixgb_driver = {
136	.name     = ixgb_driver_name,
137	.id_table = ixgb_pci_tbl,
138	.probe    = ixgb_probe,
139	.remove   = __devexit_p(ixgb_remove),
140	.err_handler = &ixgb_err_handler
141};
142
143MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
144MODULE_DESCRIPTION("Intel(R) PRO/10GbE Network Driver");
145MODULE_LICENSE("GPL");
146MODULE_VERSION(DRV_VERSION);
147
148#define DEFAULT_DEBUG_LEVEL_SHIFT 3
149static int debug = DEFAULT_DEBUG_LEVEL_SHIFT;
150module_param(debug, int, 0);
151MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
152
153/* some defines for controlling descriptor fetches in h/w */
154#define RXDCTL_WTHRESH_DEFAULT 15  /* chip writes back at this many or RXT0 */
155#define RXDCTL_PTHRESH_DEFAULT 0   /* chip considers prefech below
156                                    * this */
157#define RXDCTL_HTHRESH_DEFAULT 0   /* chip will only prefetch if tail
158                                    * is pushed this many descriptors
159                                    * from head */
160
161/**
162 * ixgb_init_module - Driver Registration Routine
163 *
164 * ixgb_init_module is the first routine called when the driver is
165 * loaded. All it does is register with the PCI subsystem.
166 **/
167
168static int __init
169ixgb_init_module(void)
170{
171	printk(KERN_INFO "%s - version %s\n",
172	       ixgb_driver_string, ixgb_driver_version);
173
174	printk(KERN_INFO "%s\n", ixgb_copyright);
175
176	return pci_register_driver(&ixgb_driver);
177}
178
179module_init(ixgb_init_module);
180
181/**
182 * ixgb_exit_module - Driver Exit Cleanup Routine
183 *
184 * ixgb_exit_module is called just before the driver is removed
185 * from memory.
186 **/
187
188static void __exit
189ixgb_exit_module(void)
190{
191	pci_unregister_driver(&ixgb_driver);
192}
193
194module_exit(ixgb_exit_module);
195
196/**
197 * ixgb_irq_disable - Mask off interrupt generation on the NIC
198 * @adapter: board private structure
199 **/
200
201static void
202ixgb_irq_disable(struct ixgb_adapter *adapter)
203{
204	atomic_inc(&adapter->irq_sem);
205	IXGB_WRITE_REG(&adapter->hw, IMC, ~0);
206	IXGB_WRITE_FLUSH(&adapter->hw);
207	synchronize_irq(adapter->pdev->irq);
208}
209
210/**
211 * ixgb_irq_enable - Enable default interrupt generation settings
212 * @adapter: board private structure
213 **/
214
215static void
216ixgb_irq_enable(struct ixgb_adapter *adapter)
217{
218	if(atomic_dec_and_test(&adapter->irq_sem)) {
219		IXGB_WRITE_REG(&adapter->hw, IMS,
220			       IXGB_INT_RXT0 | IXGB_INT_RXDMT0 | IXGB_INT_TXDW |
221			       IXGB_INT_LSC);
222		IXGB_WRITE_FLUSH(&adapter->hw);
223	}
224}
225
226int
227ixgb_up(struct ixgb_adapter *adapter)
228{
229	struct net_device *netdev = adapter->netdev;
230	int err, irq_flags = IRQF_SHARED;
231	int max_frame = netdev->mtu + ENET_HEADER_SIZE + ENET_FCS_LENGTH;
232	struct ixgb_hw *hw = &adapter->hw;
233
234	/* hardware has been reset, we need to reload some things */
235
236	ixgb_rar_set(hw, netdev->dev_addr, 0);
237	ixgb_set_multi(netdev);
238
239	ixgb_restore_vlan(adapter);
240
241	ixgb_configure_tx(adapter);
242	ixgb_setup_rctl(adapter);
243	ixgb_configure_rx(adapter);
244	ixgb_alloc_rx_buffers(adapter);
245
246	/* disable interrupts and get the hardware into a known state */
247	IXGB_WRITE_REG(&adapter->hw, IMC, 0xffffffff);
248
249	/* only enable MSI if bus is in PCI-X mode */
250	if (IXGB_READ_REG(&adapter->hw, STATUS) & IXGB_STATUS_PCIX_MODE) {
251		err = pci_enable_msi(adapter->pdev);
252		if (!err) {
253			adapter->have_msi = 1;
254			irq_flags = 0;
255		}
256		/* proceed to try to request regular interrupt */
257	}
258
259	err = request_irq(adapter->pdev->irq, &ixgb_intr, irq_flags,
260	                  netdev->name, netdev);
261	if (err) {
262		if (adapter->have_msi)
263			pci_disable_msi(adapter->pdev);
264		DPRINTK(PROBE, ERR,
265		 "Unable to allocate interrupt Error: %d\n", err);
266		return err;
267	}
268
269	if((hw->max_frame_size != max_frame) ||
270		(hw->max_frame_size !=
271		(IXGB_READ_REG(hw, MFS) >> IXGB_MFS_SHIFT))) {
272
273		hw->max_frame_size = max_frame;
274
275		IXGB_WRITE_REG(hw, MFS, hw->max_frame_size << IXGB_MFS_SHIFT);
276
277		if(hw->max_frame_size >
278		   IXGB_MAX_ENET_FRAME_SIZE_WITHOUT_FCS + ENET_FCS_LENGTH) {
279			uint32_t ctrl0 = IXGB_READ_REG(hw, CTRL0);
280
281			if(!(ctrl0 & IXGB_CTRL0_JFE)) {
282				ctrl0 |= IXGB_CTRL0_JFE;
283				IXGB_WRITE_REG(hw, CTRL0, ctrl0);
284			}
285		}
286	}
287
288	mod_timer(&adapter->watchdog_timer, jiffies);
289
290#ifdef CONFIG_IXGB_NAPI
291	netif_poll_enable(netdev);
292#endif
293	ixgb_irq_enable(adapter);
294
295	return 0;
296}
297
298void
299ixgb_down(struct ixgb_adapter *adapter, boolean_t kill_watchdog)
300{
301	struct net_device *netdev = adapter->netdev;
302
303	ixgb_irq_disable(adapter);
304	free_irq(adapter->pdev->irq, netdev);
305
306	if (adapter->have_msi)
307		pci_disable_msi(adapter->pdev);
308
309	if(kill_watchdog)
310		del_timer_sync(&adapter->watchdog_timer);
311#ifdef CONFIG_IXGB_NAPI
312	netif_poll_disable(netdev);
313#endif
314	adapter->link_speed = 0;
315	adapter->link_duplex = 0;
316	netif_carrier_off(netdev);
317	netif_stop_queue(netdev);
318
319	ixgb_reset(adapter);
320	ixgb_clean_tx_ring(adapter);
321	ixgb_clean_rx_ring(adapter);
322}
323
324void
325ixgb_reset(struct ixgb_adapter *adapter)
326{
327
328	ixgb_adapter_stop(&adapter->hw);
329	if(!ixgb_init_hw(&adapter->hw))
330		DPRINTK(PROBE, ERR, "ixgb_init_hw failed.\n");
331}
332
333/**
334 * ixgb_probe - Device Initialization Routine
335 * @pdev: PCI device information struct
336 * @ent: entry in ixgb_pci_tbl
337 *
338 * Returns 0 on success, negative on failure
339 *
340 * ixgb_probe initializes an adapter identified by a pci_dev structure.
341 * The OS initialization, configuring of the adapter private structure,
342 * and a hardware reset occur.
343 **/
344
345static int __devinit
346ixgb_probe(struct pci_dev *pdev,
347		const struct pci_device_id *ent)
348{
349	struct net_device *netdev = NULL;
350	struct ixgb_adapter *adapter;
351	static int cards_found = 0;
352	unsigned long mmio_start;
353	int mmio_len;
354	int pci_using_dac;
355	int i;
356	int err;
357
358	if((err = pci_enable_device(pdev)))
359		return err;
360
361	if(!(err = pci_set_dma_mask(pdev, DMA_64BIT_MASK)) &&
362	   !(err = pci_set_consistent_dma_mask(pdev, DMA_64BIT_MASK))) {
363		pci_using_dac = 1;
364	} else {
365		if((err = pci_set_dma_mask(pdev, DMA_32BIT_MASK)) ||
366		   (err = pci_set_consistent_dma_mask(pdev, DMA_32BIT_MASK))) {
367			printk(KERN_ERR
368			 "ixgb: No usable DMA configuration, aborting\n");
369			goto err_dma_mask;
370		}
371		pci_using_dac = 0;
372	}
373
374	if((err = pci_request_regions(pdev, ixgb_driver_name)))
375		goto err_request_regions;
376
377	pci_set_master(pdev);
378
379	netdev = alloc_etherdev(sizeof(struct ixgb_adapter));
380	if(!netdev) {
381		err = -ENOMEM;
382		goto err_alloc_etherdev;
383	}
384
385	SET_MODULE_OWNER(netdev);
386	SET_NETDEV_DEV(netdev, &pdev->dev);
387
388	pci_set_drvdata(pdev, netdev);
389	adapter = netdev_priv(netdev);
390	adapter->netdev = netdev;
391	adapter->pdev = pdev;
392	adapter->hw.back = adapter;
393	adapter->msg_enable = netif_msg_init(debug, DEFAULT_DEBUG_LEVEL_SHIFT);
394
395	mmio_start = pci_resource_start(pdev, BAR_0);
396	mmio_len = pci_resource_len(pdev, BAR_0);
397
398	adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
399	if(!adapter->hw.hw_addr) {
400		err = -EIO;
401		goto err_ioremap;
402	}
403
404	for(i = BAR_1; i <= BAR_5; i++) {
405		if(pci_resource_len(pdev, i) == 0)
406			continue;
407		if(pci_resource_flags(pdev, i) & IORESOURCE_IO) {
408			adapter->hw.io_base = pci_resource_start(pdev, i);
409			break;
410		}
411	}
412
413	netdev->open = &ixgb_open;
414	netdev->stop = &ixgb_close;
415	netdev->hard_start_xmit = &ixgb_xmit_frame;
416	netdev->get_stats = &ixgb_get_stats;
417	netdev->set_multicast_list = &ixgb_set_multi;
418	netdev->set_mac_address = &ixgb_set_mac;
419	netdev->change_mtu = &ixgb_change_mtu;
420	ixgb_set_ethtool_ops(netdev);
421	netdev->tx_timeout = &ixgb_tx_timeout;
422	netdev->watchdog_timeo = 5 * HZ;
423#ifdef CONFIG_IXGB_NAPI
424	netdev->poll = &ixgb_clean;
425	netdev->weight = 64;
426#endif
427	netdev->vlan_rx_register = ixgb_vlan_rx_register;
428	netdev->vlan_rx_add_vid = ixgb_vlan_rx_add_vid;
429	netdev->vlan_rx_kill_vid = ixgb_vlan_rx_kill_vid;
430#ifdef CONFIG_NET_POLL_CONTROLLER
431	netdev->poll_controller = ixgb_netpoll;
432#endif
433
434	strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
435	netdev->mem_start = mmio_start;
436	netdev->mem_end = mmio_start + mmio_len;
437	netdev->base_addr = adapter->hw.io_base;
438
439	adapter->bd_number = cards_found;
440	adapter->link_speed = 0;
441	adapter->link_duplex = 0;
442
443	/* setup the private structure */
444
445	if((err = ixgb_sw_init(adapter)))
446		goto err_sw_init;
447
448	netdev->features = NETIF_F_SG |
449			   NETIF_F_HW_CSUM |
450			   NETIF_F_HW_VLAN_TX |
451			   NETIF_F_HW_VLAN_RX |
452			   NETIF_F_HW_VLAN_FILTER;
453	netdev->features |= NETIF_F_TSO;
454#ifdef NETIF_F_LLTX
455	netdev->features |= NETIF_F_LLTX;
456#endif
457
458	if(pci_using_dac)
459		netdev->features |= NETIF_F_HIGHDMA;
460
461	/* make sure the EEPROM is good */
462
463	if(!ixgb_validate_eeprom_checksum(&adapter->hw)) {
464		DPRINTK(PROBE, ERR, "The EEPROM Checksum Is Not Valid\n");
465		err = -EIO;
466		goto err_eeprom;
467	}
468
469	ixgb_get_ee_mac_addr(&adapter->hw, netdev->dev_addr);
470	memcpy(netdev->perm_addr, netdev->dev_addr, netdev->addr_len);
471
472	if(!is_valid_ether_addr(netdev->perm_addr)) {
473		DPRINTK(PROBE, ERR, "Invalid MAC Address\n");
474		err = -EIO;
475		goto err_eeprom;
476	}
477
478	adapter->part_num = ixgb_get_ee_pba_number(&adapter->hw);
479
480	init_timer(&adapter->watchdog_timer);
481	adapter->watchdog_timer.function = &ixgb_watchdog;
482	adapter->watchdog_timer.data = (unsigned long)adapter;
483
484	INIT_WORK(&adapter->tx_timeout_task, ixgb_tx_timeout_task);
485
486	strcpy(netdev->name, "eth%d");
487	if((err = register_netdev(netdev)))
488		goto err_register;
489
490	/* we're going to reset, so assume we have no link for now */
491
492	netif_carrier_off(netdev);
493	netif_stop_queue(netdev);
494
495	DPRINTK(PROBE, INFO, "Intel(R) PRO/10GbE Network Connection\n");
496	ixgb_check_options(adapter);
497	/* reset the hardware with the new settings */
498
499	ixgb_reset(adapter);
500
501	cards_found++;
502	return 0;
503
504err_register:
505err_sw_init:
506err_eeprom:
507	iounmap(adapter->hw.hw_addr);
508err_ioremap:
509	free_netdev(netdev);
510err_alloc_etherdev:
511	pci_release_regions(pdev);
512err_request_regions:
513err_dma_mask:
514	pci_disable_device(pdev);
515	return err;
516}
517
518/**
519 * ixgb_remove - Device Removal Routine
520 * @pdev: PCI device information struct
521 *
522 * ixgb_remove is called by the PCI subsystem to alert the driver
523 * that it should release a PCI device.  The could be caused by a
524 * Hot-Plug event, or because the driver is going to be removed from
525 * memory.
526 **/
527
528static void __devexit
529ixgb_remove(struct pci_dev *pdev)
530{
531	struct net_device *netdev = pci_get_drvdata(pdev);
532	struct ixgb_adapter *adapter = netdev_priv(netdev);
533
534	unregister_netdev(netdev);
535
536	iounmap(adapter->hw.hw_addr);
537	pci_release_regions(pdev);
538
539	free_netdev(netdev);
540}
541
542/**
543 * ixgb_sw_init - Initialize general software structures (struct ixgb_adapter)
544 * @adapter: board private structure to initialize
545 *
546 * ixgb_sw_init initializes the Adapter private data structure.
547 * Fields are initialized based on PCI device information and
548 * OS network device settings (MTU size).
549 **/
550
551static int __devinit
552ixgb_sw_init(struct ixgb_adapter *adapter)
553{
554	struct ixgb_hw *hw = &adapter->hw;
555	struct net_device *netdev = adapter->netdev;
556	struct pci_dev *pdev = adapter->pdev;
557
558	/* PCI config space info */
559
560	hw->vendor_id = pdev->vendor;
561	hw->device_id = pdev->device;
562	hw->subsystem_vendor_id = pdev->subsystem_vendor;
563	hw->subsystem_id = pdev->subsystem_device;
564
565	hw->max_frame_size = netdev->mtu + ENET_HEADER_SIZE + ENET_FCS_LENGTH;
566	adapter->rx_buffer_len = hw->max_frame_size;
567
568	if((hw->device_id == IXGB_DEVICE_ID_82597EX)
569	   || (hw->device_id == IXGB_DEVICE_ID_82597EX_CX4)
570	   || (hw->device_id == IXGB_DEVICE_ID_82597EX_LR)
571	   || (hw->device_id == IXGB_DEVICE_ID_82597EX_SR))
572			hw->mac_type = ixgb_82597;
573	else {
574		/* should never have loaded on this device */
575		DPRINTK(PROBE, ERR, "unsupported device id\n");
576	}
577
578	/* enable flow control to be programmed */
579	hw->fc.send_xon = 1;
580
581	atomic_set(&adapter->irq_sem, 1);
582	spin_lock_init(&adapter->tx_lock);
583
584	return 0;
585}
586
587/**
588 * ixgb_open - Called when a network interface is made active
589 * @netdev: network interface device structure
590 *
591 * Returns 0 on success, negative value on failure
592 *
593 * The open entry point is called when a network interface is made
594 * active by the system (IFF_UP).  At this point all resources needed
595 * for transmit and receive operations are allocated, the interrupt
596 * handler is registered with the OS, the watchdog timer is started,
597 * and the stack is notified that the interface is ready.
598 **/
599
600static int
601ixgb_open(struct net_device *netdev)
602{
603	struct ixgb_adapter *adapter = netdev_priv(netdev);
604	int err;
605
606	/* allocate transmit descriptors */
607
608	if((err = ixgb_setup_tx_resources(adapter)))
609		goto err_setup_tx;
610
611	/* allocate receive descriptors */
612
613	if((err = ixgb_setup_rx_resources(adapter)))
614		goto err_setup_rx;
615
616	if((err = ixgb_up(adapter)))
617		goto err_up;
618
619	return 0;
620
621err_up:
622	ixgb_free_rx_resources(adapter);
623err_setup_rx:
624	ixgb_free_tx_resources(adapter);
625err_setup_tx:
626	ixgb_reset(adapter);
627
628	return err;
629}
630
631/**
632 * ixgb_close - Disables a network interface
633 * @netdev: network interface device structure
634 *
635 * Returns 0, this is not allowed to fail
636 *
637 * The close entry point is called when an interface is de-activated
638 * by the OS.  The hardware is still under the drivers control, but
639 * needs to be disabled.  A global MAC reset is issued to stop the
640 * hardware, and all transmit and receive resources are freed.
641 **/
642
643static int
644ixgb_close(struct net_device *netdev)
645{
646	struct ixgb_adapter *adapter = netdev_priv(netdev);
647
648	ixgb_down(adapter, TRUE);
649
650	ixgb_free_tx_resources(adapter);
651	ixgb_free_rx_resources(adapter);
652
653	return 0;
654}
655
656/**
657 * ixgb_setup_tx_resources - allocate Tx resources (Descriptors)
658 * @adapter: board private structure
659 *
660 * Return 0 on success, negative on failure
661 **/
662
663int
664ixgb_setup_tx_resources(struct ixgb_adapter *adapter)
665{
666	struct ixgb_desc_ring *txdr = &adapter->tx_ring;
667	struct pci_dev *pdev = adapter->pdev;
668	int size;
669
670	size = sizeof(struct ixgb_buffer) * txdr->count;
671	txdr->buffer_info = vmalloc(size);
672	if(!txdr->buffer_info) {
673		DPRINTK(PROBE, ERR,
674		 "Unable to allocate transmit descriptor ring memory\n");
675		return -ENOMEM;
676	}
677	memset(txdr->buffer_info, 0, size);
678
679	/* round up to nearest 4K */
680
681	txdr->size = txdr->count * sizeof(struct ixgb_tx_desc);
682	txdr->size = ALIGN(txdr->size, 4096);
683
684	txdr->desc = pci_alloc_consistent(pdev, txdr->size, &txdr->dma);
685	if(!txdr->desc) {
686		vfree(txdr->buffer_info);
687		DPRINTK(PROBE, ERR,
688		 "Unable to allocate transmit descriptor memory\n");
689		return -ENOMEM;
690	}
691	memset(txdr->desc, 0, txdr->size);
692
693	txdr->next_to_use = 0;
694	txdr->next_to_clean = 0;
695
696	return 0;
697}
698
699/**
700 * ixgb_configure_tx - Configure 82597 Transmit Unit after Reset.
701 * @adapter: board private structure
702 *
703 * Configure the Tx unit of the MAC after a reset.
704 **/
705
706static void
707ixgb_configure_tx(struct ixgb_adapter *adapter)
708{
709	uint64_t tdba = adapter->tx_ring.dma;
710	uint32_t tdlen = adapter->tx_ring.count * sizeof(struct ixgb_tx_desc);
711	uint32_t tctl;
712	struct ixgb_hw *hw = &adapter->hw;
713
714	/* Setup the Base and Length of the Tx Descriptor Ring
715	 * tx_ring.dma can be either a 32 or 64 bit value
716	 */
717
718	IXGB_WRITE_REG(hw, TDBAL, (tdba & 0x00000000ffffffffULL));
719	IXGB_WRITE_REG(hw, TDBAH, (tdba >> 32));
720
721	IXGB_WRITE_REG(hw, TDLEN, tdlen);
722
723	/* Setup the HW Tx Head and Tail descriptor pointers */
724
725	IXGB_WRITE_REG(hw, TDH, 0);
726	IXGB_WRITE_REG(hw, TDT, 0);
727
728	/* don't set up txdctl, it induces performance problems if configured
729	 * incorrectly */
730	/* Set the Tx Interrupt Delay register */
731
732	IXGB_WRITE_REG(hw, TIDV, adapter->tx_int_delay);
733
734	/* Program the Transmit Control Register */
735
736	tctl = IXGB_TCTL_TCE | IXGB_TCTL_TXEN | IXGB_TCTL_TPDE;
737	IXGB_WRITE_REG(hw, TCTL, tctl);
738
739	/* Setup Transmit Descriptor Settings for this adapter */
740	adapter->tx_cmd_type =
741		IXGB_TX_DESC_TYPE
742		| (adapter->tx_int_delay_enable ? IXGB_TX_DESC_CMD_IDE : 0);
743}
744
745/**
746 * ixgb_setup_rx_resources - allocate Rx resources (Descriptors)
747 * @adapter: board private structure
748 *
749 * Returns 0 on success, negative on failure
750 **/
751
752int
753ixgb_setup_rx_resources(struct ixgb_adapter *adapter)
754{
755	struct ixgb_desc_ring *rxdr = &adapter->rx_ring;
756	struct pci_dev *pdev = adapter->pdev;
757	int size;
758
759	size = sizeof(struct ixgb_buffer) * rxdr->count;
760	rxdr->buffer_info = vmalloc(size);
761	if(!rxdr->buffer_info) {
762		DPRINTK(PROBE, ERR,
763		 "Unable to allocate receive descriptor ring\n");
764		return -ENOMEM;
765	}
766	memset(rxdr->buffer_info, 0, size);
767
768	/* Round up to nearest 4K */
769
770	rxdr->size = rxdr->count * sizeof(struct ixgb_rx_desc);
771	rxdr->size = ALIGN(rxdr->size, 4096);
772
773	rxdr->desc = pci_alloc_consistent(pdev, rxdr->size, &rxdr->dma);
774
775	if(!rxdr->desc) {
776		vfree(rxdr->buffer_info);
777		DPRINTK(PROBE, ERR,
778		 "Unable to allocate receive descriptors\n");
779		return -ENOMEM;
780	}
781	memset(rxdr->desc, 0, rxdr->size);
782
783	rxdr->next_to_clean = 0;
784	rxdr->next_to_use = 0;
785
786	return 0;
787}
788
789/**
790 * ixgb_setup_rctl - configure the receive control register
791 * @adapter: Board private structure
792 **/
793
794static void
795ixgb_setup_rctl(struct ixgb_adapter *adapter)
796{
797	uint32_t rctl;
798
799	rctl = IXGB_READ_REG(&adapter->hw, RCTL);
800
801	rctl &= ~(3 << IXGB_RCTL_MO_SHIFT);
802
803	rctl |=
804		IXGB_RCTL_BAM | IXGB_RCTL_RDMTS_1_2 |
805		IXGB_RCTL_RXEN | IXGB_RCTL_CFF |
806		(adapter->hw.mc_filter_type << IXGB_RCTL_MO_SHIFT);
807
808	rctl |= IXGB_RCTL_SECRC;
809
810	if (adapter->rx_buffer_len <= IXGB_RXBUFFER_2048)
811		rctl |= IXGB_RCTL_BSIZE_2048;
812	else if (adapter->rx_buffer_len <= IXGB_RXBUFFER_4096)
813		rctl |= IXGB_RCTL_BSIZE_4096;
814	else if (adapter->rx_buffer_len <= IXGB_RXBUFFER_8192)
815		rctl |= IXGB_RCTL_BSIZE_8192;
816	else if (adapter->rx_buffer_len <= IXGB_RXBUFFER_16384)
817		rctl |= IXGB_RCTL_BSIZE_16384;
818
819	IXGB_WRITE_REG(&adapter->hw, RCTL, rctl);
820}
821
822/**
823 * ixgb_configure_rx - Configure 82597 Receive Unit after Reset.
824 * @adapter: board private structure
825 *
826 * Configure the Rx unit of the MAC after a reset.
827 **/
828
829static void
830ixgb_configure_rx(struct ixgb_adapter *adapter)
831{
832	uint64_t rdba = adapter->rx_ring.dma;
833	uint32_t rdlen = adapter->rx_ring.count * sizeof(struct ixgb_rx_desc);
834	struct ixgb_hw *hw = &adapter->hw;
835	uint32_t rctl;
836	uint32_t rxcsum;
837	uint32_t rxdctl;
838
839	/* make sure receives are disabled while setting up the descriptors */
840
841	rctl = IXGB_READ_REG(hw, RCTL);
842	IXGB_WRITE_REG(hw, RCTL, rctl & ~IXGB_RCTL_RXEN);
843
844	/* set the Receive Delay Timer Register */
845
846	IXGB_WRITE_REG(hw, RDTR, adapter->rx_int_delay);
847
848	/* Setup the Base and Length of the Rx Descriptor Ring */
849
850	IXGB_WRITE_REG(hw, RDBAL, (rdba & 0x00000000ffffffffULL));
851	IXGB_WRITE_REG(hw, RDBAH, (rdba >> 32));
852
853	IXGB_WRITE_REG(hw, RDLEN, rdlen);
854
855	/* Setup the HW Rx Head and Tail Descriptor Pointers */
856	IXGB_WRITE_REG(hw, RDH, 0);
857	IXGB_WRITE_REG(hw, RDT, 0);
858
859	/* set up pre-fetching of receive buffers so we get some before we
860	 * run out (default hardware behavior is to run out before fetching
861	 * more).  This sets up to fetch if HTHRESH rx descriptors are avail
862	 * and the descriptors in hw cache are below PTHRESH.  This avoids
863	 * the hardware behavior of fetching <=512 descriptors in a single
864	 * burst that pre-empts all other activity, usually causing fifo
865	 * overflows. */
866	/* use WTHRESH to burst write 16 descriptors or burst when RXT0 */
867	rxdctl = RXDCTL_WTHRESH_DEFAULT << IXGB_RXDCTL_WTHRESH_SHIFT |
868	         RXDCTL_HTHRESH_DEFAULT << IXGB_RXDCTL_HTHRESH_SHIFT |
869	         RXDCTL_PTHRESH_DEFAULT << IXGB_RXDCTL_PTHRESH_SHIFT;
870	IXGB_WRITE_REG(hw, RXDCTL, rxdctl);
871
872	/* Enable Receive Checksum Offload for TCP and UDP */
873	if(adapter->rx_csum == TRUE) {
874		rxcsum = IXGB_READ_REG(hw, RXCSUM);
875		rxcsum |= IXGB_RXCSUM_TUOFL;
876		IXGB_WRITE_REG(hw, RXCSUM, rxcsum);
877	}
878
879	/* Enable Receives */
880
881	IXGB_WRITE_REG(hw, RCTL, rctl);
882}
883
884/**
885 * ixgb_free_tx_resources - Free Tx Resources
886 * @adapter: board private structure
887 *
888 * Free all transmit software resources
889 **/
890
891void
892ixgb_free_tx_resources(struct ixgb_adapter *adapter)
893{
894	struct pci_dev *pdev = adapter->pdev;
895
896	ixgb_clean_tx_ring(adapter);
897
898	vfree(adapter->tx_ring.buffer_info);
899	adapter->tx_ring.buffer_info = NULL;
900
901	pci_free_consistent(pdev, adapter->tx_ring.size,
902			    adapter->tx_ring.desc, adapter->tx_ring.dma);
903
904	adapter->tx_ring.desc = NULL;
905}
906
907static void
908ixgb_unmap_and_free_tx_resource(struct ixgb_adapter *adapter,
909					struct ixgb_buffer *buffer_info)
910{
911	struct pci_dev *pdev = adapter->pdev;
912
913	if (buffer_info->dma)
914		pci_unmap_page(pdev, buffer_info->dma, buffer_info->length,
915		               PCI_DMA_TODEVICE);
916
917	if (buffer_info->skb)
918		dev_kfree_skb_any(buffer_info->skb);
919
920	buffer_info->skb = NULL;
921	buffer_info->dma = 0;
922	buffer_info->time_stamp = 0;
923	/* these fields must always be initialized in tx
924	 * buffer_info->length = 0;
925	 * buffer_info->next_to_watch = 0; */
926}
927
928/**
929 * ixgb_clean_tx_ring - Free Tx Buffers
930 * @adapter: board private structure
931 **/
932
933static void
934ixgb_clean_tx_ring(struct ixgb_adapter *adapter)
935{
936	struct ixgb_desc_ring *tx_ring = &adapter->tx_ring;
937	struct ixgb_buffer *buffer_info;
938	unsigned long size;
939	unsigned int i;
940
941	/* Free all the Tx ring sk_buffs */
942
943	for(i = 0; i < tx_ring->count; i++) {
944		buffer_info = &tx_ring->buffer_info[i];
945		ixgb_unmap_and_free_tx_resource(adapter, buffer_info);
946	}
947
948	size = sizeof(struct ixgb_buffer) * tx_ring->count;
949	memset(tx_ring->buffer_info, 0, size);
950
951	/* Zero out the descriptor ring */
952
953	memset(tx_ring->desc, 0, tx_ring->size);
954
955	tx_ring->next_to_use = 0;
956	tx_ring->next_to_clean = 0;
957
958	IXGB_WRITE_REG(&adapter->hw, TDH, 0);
959	IXGB_WRITE_REG(&adapter->hw, TDT, 0);
960}
961
962/**
963 * ixgb_free_rx_resources - Free Rx Resources
964 * @adapter: board private structure
965 *
966 * Free all receive software resources
967 **/
968
969void
970ixgb_free_rx_resources(struct ixgb_adapter *adapter)
971{
972	struct ixgb_desc_ring *rx_ring = &adapter->rx_ring;
973	struct pci_dev *pdev = adapter->pdev;
974
975	ixgb_clean_rx_ring(adapter);
976
977	vfree(rx_ring->buffer_info);
978	rx_ring->buffer_info = NULL;
979
980	pci_free_consistent(pdev, rx_ring->size, rx_ring->desc, rx_ring->dma);
981
982	rx_ring->desc = NULL;
983}
984
985/**
986 * ixgb_clean_rx_ring - Free Rx Buffers
987 * @adapter: board private structure
988 **/
989
990static void
991ixgb_clean_rx_ring(struct ixgb_adapter *adapter)
992{
993	struct ixgb_desc_ring *rx_ring = &adapter->rx_ring;
994	struct ixgb_buffer *buffer_info;
995	struct pci_dev *pdev = adapter->pdev;
996	unsigned long size;
997	unsigned int i;
998
999	/* Free all the Rx ring sk_buffs */
1000
1001	for(i = 0; i < rx_ring->count; i++) {
1002		buffer_info = &rx_ring->buffer_info[i];
1003		if(buffer_info->skb) {
1004
1005			pci_unmap_single(pdev,
1006					 buffer_info->dma,
1007					 buffer_info->length,
1008					 PCI_DMA_FROMDEVICE);
1009
1010			dev_kfree_skb(buffer_info->skb);
1011
1012			buffer_info->skb = NULL;
1013		}
1014	}
1015
1016	size = sizeof(struct ixgb_buffer) * rx_ring->count;
1017	memset(rx_ring->buffer_info, 0, size);
1018
1019	/* Zero out the descriptor ring */
1020
1021	memset(rx_ring->desc, 0, rx_ring->size);
1022
1023	rx_ring->next_to_clean = 0;
1024	rx_ring->next_to_use = 0;
1025
1026	IXGB_WRITE_REG(&adapter->hw, RDH, 0);
1027	IXGB_WRITE_REG(&adapter->hw, RDT, 0);
1028}
1029
1030/**
1031 * ixgb_set_mac - Change the Ethernet Address of the NIC
1032 * @netdev: network interface device structure
1033 * @p: pointer to an address structure
1034 *
1035 * Returns 0 on success, negative on failure
1036 **/
1037
1038static int
1039ixgb_set_mac(struct net_device *netdev, void *p)
1040{
1041	struct ixgb_adapter *adapter = netdev_priv(netdev);
1042	struct sockaddr *addr = p;
1043
1044	if(!is_valid_ether_addr(addr->sa_data))
1045		return -EADDRNOTAVAIL;
1046
1047	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
1048
1049	ixgb_rar_set(&adapter->hw, addr->sa_data, 0);
1050
1051	return 0;
1052}
1053
1054/**
1055 * ixgb_set_multi - Multicast and Promiscuous mode set
1056 * @netdev: network interface device structure
1057 *
1058 * The set_multi entry point is called whenever the multicast address
1059 * list or the network interface flags are updated.  This routine is
1060 * responsible for configuring the hardware for proper multicast,
1061 * promiscuous mode, and all-multi behavior.
1062 **/
1063
1064static void
1065ixgb_set_multi(struct net_device *netdev)
1066{
1067	struct ixgb_adapter *adapter = netdev_priv(netdev);
1068	struct ixgb_hw *hw = &adapter->hw;
1069	struct dev_mc_list *mc_ptr;
1070	uint32_t rctl;
1071	int i;
1072
1073	/* Check for Promiscuous and All Multicast modes */
1074
1075	rctl = IXGB_READ_REG(hw, RCTL);
1076
1077	if(netdev->flags & IFF_PROMISC) {
1078		rctl |= (IXGB_RCTL_UPE | IXGB_RCTL_MPE);
1079	} else if(netdev->flags & IFF_ALLMULTI) {
1080		rctl |= IXGB_RCTL_MPE;
1081		rctl &= ~IXGB_RCTL_UPE;
1082	} else {
1083		rctl &= ~(IXGB_RCTL_UPE | IXGB_RCTL_MPE);
1084	}
1085
1086	if(netdev->mc_count > IXGB_MAX_NUM_MULTICAST_ADDRESSES) {
1087		rctl |= IXGB_RCTL_MPE;
1088		IXGB_WRITE_REG(hw, RCTL, rctl);
1089	} else {
1090		uint8_t mta[netdev->mc_count * IXGB_ETH_LENGTH_OF_ADDRESS];
1091
1092		IXGB_WRITE_REG(hw, RCTL, rctl);
1093
1094		for(i = 0, mc_ptr = netdev->mc_list; mc_ptr;
1095			i++, mc_ptr = mc_ptr->next)
1096			memcpy(&mta[i * IXGB_ETH_LENGTH_OF_ADDRESS],
1097				   mc_ptr->dmi_addr, IXGB_ETH_LENGTH_OF_ADDRESS);
1098
1099		ixgb_mc_addr_list_update(hw, mta, netdev->mc_count, 0);
1100	}
1101}
1102
1103/**
1104 * ixgb_watchdog - Timer Call-back
1105 * @data: pointer to netdev cast into an unsigned long
1106 **/
1107
1108static void
1109ixgb_watchdog(unsigned long data)
1110{
1111	struct ixgb_adapter *adapter = (struct ixgb_adapter *)data;
1112	struct net_device *netdev = adapter->netdev;
1113	struct ixgb_desc_ring *txdr = &adapter->tx_ring;
1114
1115	ixgb_check_for_link(&adapter->hw);
1116
1117	if (ixgb_check_for_bad_link(&adapter->hw)) {
1118		/* force the reset path */
1119		netif_stop_queue(netdev);
1120	}
1121
1122	if(adapter->hw.link_up) {
1123		if(!netif_carrier_ok(netdev)) {
1124			DPRINTK(LINK, INFO,
1125			        "NIC Link is Up 10000 Mbps Full Duplex\n");
1126			adapter->link_speed = 10000;
1127			adapter->link_duplex = FULL_DUPLEX;
1128			netif_carrier_on(netdev);
1129			netif_wake_queue(netdev);
1130		}
1131	} else {
1132		if(netif_carrier_ok(netdev)) {
1133			adapter->link_speed = 0;
1134			adapter->link_duplex = 0;
1135			DPRINTK(LINK, INFO, "NIC Link is Down\n");
1136			netif_carrier_off(netdev);
1137			netif_stop_queue(netdev);
1138
1139		}
1140	}
1141
1142	ixgb_update_stats(adapter);
1143
1144	if(!netif_carrier_ok(netdev)) {
1145		if(IXGB_DESC_UNUSED(txdr) + 1 < txdr->count) {
1146			/* We've lost link, so the controller stops DMA,
1147			 * but we've got queued Tx work that's never going
1148			 * to get done, so reset controller to flush Tx.
1149			 * (Do the reset outside of interrupt context). */
1150			schedule_work(&adapter->tx_timeout_task);
1151		}
1152	}
1153
1154	/* Force detection of hung controller every watchdog period */
1155	adapter->detect_tx_hung = TRUE;
1156
1157	/* generate an interrupt to force clean up of any stragglers */
1158	IXGB_WRITE_REG(&adapter->hw, ICS, IXGB_INT_TXDW);
1159
1160	/* Reset the timer */
1161	mod_timer(&adapter->watchdog_timer, jiffies + 2 * HZ);
1162}
1163
1164#define IXGB_TX_FLAGS_CSUM		0x00000001
1165#define IXGB_TX_FLAGS_VLAN		0x00000002
1166#define IXGB_TX_FLAGS_TSO		0x00000004
1167
1168static int
1169ixgb_tso(struct ixgb_adapter *adapter, struct sk_buff *skb)
1170{
1171	struct ixgb_context_desc *context_desc;
1172	unsigned int i;
1173	uint8_t ipcss, ipcso, tucss, tucso, hdr_len;
1174	uint16_t ipcse, tucse, mss;
1175	int err;
1176
1177	if (likely(skb_is_gso(skb))) {
1178		struct ixgb_buffer *buffer_info;
1179		struct iphdr *iph;
1180
1181		if (skb_header_cloned(skb)) {
1182			err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
1183			if (err)
1184				return err;
1185		}
1186
1187		hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
1188		mss = skb_shinfo(skb)->gso_size;
1189		iph = ip_hdr(skb);
1190		iph->tot_len = 0;
1191		iph->check = 0;
1192		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
1193							 iph->daddr, 0,
1194							 IPPROTO_TCP, 0);
1195		ipcss = skb_network_offset(skb);
1196		ipcso = (void *)&(iph->check) - (void *)skb->data;
1197		ipcse = skb_transport_offset(skb) - 1;
1198		tucss = skb_transport_offset(skb);
1199		tucso = (void *)&(tcp_hdr(skb)->check) - (void *)skb->data;
1200		tucse = 0;
1201
1202		i = adapter->tx_ring.next_to_use;
1203		context_desc = IXGB_CONTEXT_DESC(adapter->tx_ring, i);
1204		buffer_info = &adapter->tx_ring.buffer_info[i];
1205		WARN_ON(buffer_info->dma != 0);
1206
1207		context_desc->ipcss = ipcss;
1208		context_desc->ipcso = ipcso;
1209		context_desc->ipcse = cpu_to_le16(ipcse);
1210		context_desc->tucss = tucss;
1211		context_desc->tucso = tucso;
1212		context_desc->tucse = cpu_to_le16(tucse);
1213		context_desc->mss = cpu_to_le16(mss);
1214		context_desc->hdr_len = hdr_len;
1215		context_desc->status = 0;
1216		context_desc->cmd_type_len = cpu_to_le32(
1217						  IXGB_CONTEXT_DESC_TYPE
1218						| IXGB_CONTEXT_DESC_CMD_TSE
1219						| IXGB_CONTEXT_DESC_CMD_IP
1220						| IXGB_CONTEXT_DESC_CMD_TCP
1221						| IXGB_CONTEXT_DESC_CMD_IDE
1222						| (skb->len - (hdr_len)));
1223
1224
1225		if(++i == adapter->tx_ring.count) i = 0;
1226		adapter->tx_ring.next_to_use = i;
1227
1228		return 1;
1229	}
1230
1231	return 0;
1232}
1233
1234static boolean_t
1235ixgb_tx_csum(struct ixgb_adapter *adapter, struct sk_buff *skb)
1236{
1237	struct ixgb_context_desc *context_desc;
1238	unsigned int i;
1239	uint8_t css, cso;
1240
1241	if(likely(skb->ip_summed == CHECKSUM_PARTIAL)) {
1242		struct ixgb_buffer *buffer_info;
1243		css = skb_transport_offset(skb);
1244		cso = css + skb->csum_offset;
1245
1246		i = adapter->tx_ring.next_to_use;
1247		context_desc = IXGB_CONTEXT_DESC(adapter->tx_ring, i);
1248		buffer_info = &adapter->tx_ring.buffer_info[i];
1249		WARN_ON(buffer_info->dma != 0);
1250
1251		context_desc->tucss = css;
1252		context_desc->tucso = cso;
1253		context_desc->tucse = 0;
1254		/* zero out any previously existing data in one instruction */
1255		*(uint32_t *)&(context_desc->ipcss) = 0;
1256		context_desc->status = 0;
1257		context_desc->hdr_len = 0;
1258		context_desc->mss = 0;
1259		context_desc->cmd_type_len =
1260			cpu_to_le32(IXGB_CONTEXT_DESC_TYPE
1261				    | IXGB_TX_DESC_CMD_IDE);
1262
1263		if(++i == adapter->tx_ring.count) i = 0;
1264		adapter->tx_ring.next_to_use = i;
1265
1266		return TRUE;
1267	}
1268
1269	return FALSE;
1270}
1271
1272#define IXGB_MAX_TXD_PWR	14
1273#define IXGB_MAX_DATA_PER_TXD	(1<<IXGB_MAX_TXD_PWR)
1274
1275static int
1276ixgb_tx_map(struct ixgb_adapter *adapter, struct sk_buff *skb,
1277	    unsigned int first)
1278{
1279	struct ixgb_desc_ring *tx_ring = &adapter->tx_ring;
1280	struct ixgb_buffer *buffer_info;
1281	int len = skb->len;
1282	unsigned int offset = 0, size, count = 0, i;
1283	unsigned int mss = skb_shinfo(skb)->gso_size;
1284
1285	unsigned int nr_frags = skb_shinfo(skb)->nr_frags;
1286	unsigned int f;
1287
1288	len -= skb->data_len;
1289
1290	i = tx_ring->next_to_use;
1291
1292	while(len) {
1293		buffer_info = &tx_ring->buffer_info[i];
1294		size = min(len, IXGB_MAX_DATA_PER_TXD);
1295		if (unlikely(mss && !nr_frags && size == len && size > 8))
1296			size -= 4;
1297
1298		buffer_info->length = size;
1299		WARN_ON(buffer_info->dma != 0);
1300		buffer_info->dma =
1301			pci_map_single(adapter->pdev,
1302				skb->data + offset,
1303				size,
1304				PCI_DMA_TODEVICE);
1305		buffer_info->time_stamp = jiffies;
1306		buffer_info->next_to_watch = 0;
1307
1308		len -= size;
1309		offset += size;
1310		count++;
1311		if(++i == tx_ring->count) i = 0;
1312	}
1313
1314	for(f = 0; f < nr_frags; f++) {
1315		struct skb_frag_struct *frag;
1316
1317		frag = &skb_shinfo(skb)->frags[f];
1318		len = frag->size;
1319		offset = 0;
1320
1321		while(len) {
1322			buffer_info = &tx_ring->buffer_info[i];
1323			size = min(len, IXGB_MAX_DATA_PER_TXD);
1324
1325			if (unlikely(mss && !nr_frags && size == len
1326			             && size > 8))
1327				size -= 4;
1328
1329			buffer_info->length = size;
1330			buffer_info->dma =
1331				pci_map_page(adapter->pdev,
1332					frag->page,
1333					frag->page_offset + offset,
1334					size,
1335					PCI_DMA_TODEVICE);
1336			buffer_info->time_stamp = jiffies;
1337			buffer_info->next_to_watch = 0;
1338
1339			len -= size;
1340			offset += size;
1341			count++;
1342			if(++i == tx_ring->count) i = 0;
1343		}
1344	}
1345	i = (i == 0) ? tx_ring->count - 1 : i - 1;
1346	tx_ring->buffer_info[i].skb = skb;
1347	tx_ring->buffer_info[first].next_to_watch = i;
1348
1349	return count;
1350}
1351
1352static void
1353ixgb_tx_queue(struct ixgb_adapter *adapter, int count, int vlan_id,int tx_flags)
1354{
1355	struct ixgb_desc_ring *tx_ring = &adapter->tx_ring;
1356	struct ixgb_tx_desc *tx_desc = NULL;
1357	struct ixgb_buffer *buffer_info;
1358	uint32_t cmd_type_len = adapter->tx_cmd_type;
1359	uint8_t status = 0;
1360	uint8_t popts = 0;
1361	unsigned int i;
1362
1363	if(tx_flags & IXGB_TX_FLAGS_TSO) {
1364		cmd_type_len |= IXGB_TX_DESC_CMD_TSE;
1365		popts |= (IXGB_TX_DESC_POPTS_IXSM | IXGB_TX_DESC_POPTS_TXSM);
1366	}
1367
1368	if(tx_flags & IXGB_TX_FLAGS_CSUM)
1369		popts |= IXGB_TX_DESC_POPTS_TXSM;
1370
1371	if(tx_flags & IXGB_TX_FLAGS_VLAN) {
1372		cmd_type_len |= IXGB_TX_DESC_CMD_VLE;
1373	}
1374
1375	i = tx_ring->next_to_use;
1376
1377	while(count--) {
1378		buffer_info = &tx_ring->buffer_info[i];
1379		tx_desc = IXGB_TX_DESC(*tx_ring, i);
1380		tx_desc->buff_addr = cpu_to_le64(buffer_info->dma);
1381		tx_desc->cmd_type_len =
1382			cpu_to_le32(cmd_type_len | buffer_info->length);
1383		tx_desc->status = status;
1384		tx_desc->popts = popts;
1385		tx_desc->vlan = cpu_to_le16(vlan_id);
1386
1387		if(++i == tx_ring->count) i = 0;
1388	}
1389
1390	tx_desc->cmd_type_len |= cpu_to_le32(IXGB_TX_DESC_CMD_EOP
1391				| IXGB_TX_DESC_CMD_RS );
1392
1393	/* Force memory writes to complete before letting h/w
1394	 * know there are new descriptors to fetch.  (Only
1395	 * applicable for weak-ordered memory model archs,
1396	 * such as IA-64). */
1397	wmb();
1398
1399	tx_ring->next_to_use = i;
1400	IXGB_WRITE_REG(&adapter->hw, TDT, i);
1401}
1402
1403static int __ixgb_maybe_stop_tx(struct net_device *netdev, int size)
1404{
1405	struct ixgb_adapter *adapter = netdev_priv(netdev);
1406	struct ixgb_desc_ring *tx_ring = &adapter->tx_ring;
1407
1408	netif_stop_queue(netdev);
1409	/* Herbert's original patch had:
1410	 *  smp_mb__after_netif_stop_queue();
1411	 * but since that doesn't exist yet, just open code it. */
1412	smp_mb();
1413
1414	/* We need to check again in a case another CPU has just
1415	 * made room available. */
1416	if (likely(IXGB_DESC_UNUSED(tx_ring) < size))
1417		return -EBUSY;
1418
1419	/* A reprieve! */
1420	netif_start_queue(netdev);
1421	++adapter->restart_queue;
1422	return 0;
1423}
1424
1425static int ixgb_maybe_stop_tx(struct net_device *netdev,
1426                              struct ixgb_desc_ring *tx_ring, int size)
1427{
1428	if (likely(IXGB_DESC_UNUSED(tx_ring) >= size))
1429		return 0;
1430	return __ixgb_maybe_stop_tx(netdev, size);
1431}
1432
1433
1434/* Tx Descriptors needed, worst case */
1435#define TXD_USE_COUNT(S) (((S) >> IXGB_MAX_TXD_PWR) + \
1436			 (((S) & (IXGB_MAX_DATA_PER_TXD - 1)) ? 1 : 0))
1437#define DESC_NEEDED TXD_USE_COUNT(IXGB_MAX_DATA_PER_TXD) /* skb->date */ + \
1438	MAX_SKB_FRAGS * TXD_USE_COUNT(PAGE_SIZE) + 1 /* for context */ \
1439	+ 1
1440
1441static int
1442ixgb_xmit_frame(struct sk_buff *skb, struct net_device *netdev)
1443{
1444	struct ixgb_adapter *adapter = netdev_priv(netdev);
1445	unsigned int first;
1446	unsigned int tx_flags = 0;
1447	unsigned long flags;
1448	int vlan_id = 0;
1449	int tso;
1450
1451	if(skb->len <= 0) {
1452		dev_kfree_skb_any(skb);
1453		return 0;
1454	}
1455
1456#ifdef NETIF_F_LLTX
1457	local_irq_save(flags);
1458	if (!spin_trylock(&adapter->tx_lock)) {
1459		/* Collision - tell upper layer to requeue */
1460		local_irq_restore(flags);
1461		return NETDEV_TX_LOCKED;
1462	}
1463#else
1464	spin_lock_irqsave(&adapter->tx_lock, flags);
1465#endif
1466
1467	if (unlikely(ixgb_maybe_stop_tx(netdev, &adapter->tx_ring,
1468                     DESC_NEEDED))) {
1469		netif_stop_queue(netdev);
1470		spin_unlock_irqrestore(&adapter->tx_lock, flags);
1471		return NETDEV_TX_BUSY;
1472	}
1473
1474#ifndef NETIF_F_LLTX
1475	spin_unlock_irqrestore(&adapter->tx_lock, flags);
1476#endif
1477
1478	if(adapter->vlgrp && vlan_tx_tag_present(skb)) {
1479		tx_flags |= IXGB_TX_FLAGS_VLAN;
1480		vlan_id = vlan_tx_tag_get(skb);
1481	}
1482
1483	first = adapter->tx_ring.next_to_use;
1484
1485	tso = ixgb_tso(adapter, skb);
1486	if (tso < 0) {
1487		dev_kfree_skb_any(skb);
1488#ifdef NETIF_F_LLTX
1489		spin_unlock_irqrestore(&adapter->tx_lock, flags);
1490#endif
1491		return NETDEV_TX_OK;
1492	}
1493
1494	if (likely(tso))
1495		tx_flags |= IXGB_TX_FLAGS_TSO;
1496	else if(ixgb_tx_csum(adapter, skb))
1497		tx_flags |= IXGB_TX_FLAGS_CSUM;
1498
1499	ixgb_tx_queue(adapter, ixgb_tx_map(adapter, skb, first), vlan_id,
1500			tx_flags);
1501
1502	netdev->trans_start = jiffies;
1503
1504#ifdef NETIF_F_LLTX
1505	/* Make sure there is space in the ring for the next send. */
1506	ixgb_maybe_stop_tx(netdev, &adapter->tx_ring, DESC_NEEDED);
1507
1508	spin_unlock_irqrestore(&adapter->tx_lock, flags);
1509
1510#endif
1511	return NETDEV_TX_OK;
1512}
1513
1514/**
1515 * ixgb_tx_timeout - Respond to a Tx Hang
1516 * @netdev: network interface device structure
1517 **/
1518
1519static void
1520ixgb_tx_timeout(struct net_device *netdev)
1521{
1522	struct ixgb_adapter *adapter = netdev_priv(netdev);
1523
1524	/* Do the reset outside of interrupt context */
1525	schedule_work(&adapter->tx_timeout_task);
1526}
1527
1528static void
1529ixgb_tx_timeout_task(struct work_struct *work)
1530{
1531	struct ixgb_adapter *adapter =
1532		container_of(work, struct ixgb_adapter, tx_timeout_task);
1533
1534	adapter->tx_timeout_count++;
1535	ixgb_down(adapter, TRUE);
1536	ixgb_up(adapter);
1537}
1538
1539/**
1540 * ixgb_get_stats - Get System Network Statistics
1541 * @netdev: network interface device structure
1542 *
1543 * Returns the address of the device statistics structure.
1544 * The statistics are actually updated from the timer callback.
1545 **/
1546
1547static struct net_device_stats *
1548ixgb_get_stats(struct net_device *netdev)
1549{
1550	struct ixgb_adapter *adapter = netdev_priv(netdev);
1551
1552	return &adapter->net_stats;
1553}
1554
1555/**
1556 * ixgb_change_mtu - Change the Maximum Transfer Unit
1557 * @netdev: network interface device structure
1558 * @new_mtu: new value for maximum frame size
1559 *
1560 * Returns 0 on success, negative on failure
1561 **/
1562
1563static int
1564ixgb_change_mtu(struct net_device *netdev, int new_mtu)
1565{
1566	struct ixgb_adapter *adapter = netdev_priv(netdev);
1567	int max_frame = new_mtu + ENET_HEADER_SIZE + ENET_FCS_LENGTH;
1568	int old_max_frame = netdev->mtu + ENET_HEADER_SIZE + ENET_FCS_LENGTH;
1569
1570
1571	if((max_frame < IXGB_MIN_ENET_FRAME_SIZE_WITHOUT_FCS + ENET_FCS_LENGTH)
1572	   || (max_frame > IXGB_MAX_JUMBO_FRAME_SIZE + ENET_FCS_LENGTH)) {
1573		DPRINTK(PROBE, ERR, "Invalid MTU setting %d\n", new_mtu);
1574		return -EINVAL;
1575	}
1576
1577	adapter->rx_buffer_len = max_frame;
1578
1579	netdev->mtu = new_mtu;
1580
1581	if ((old_max_frame != max_frame) && netif_running(netdev)) {
1582		ixgb_down(adapter, TRUE);
1583		ixgb_up(adapter);
1584	}
1585
1586	return 0;
1587}
1588
1589/**
1590 * ixgb_update_stats - Update the board statistics counters.
1591 * @adapter: board private structure
1592 **/
1593
1594void
1595ixgb_update_stats(struct ixgb_adapter *adapter)
1596{
1597	struct net_device *netdev = adapter->netdev;
1598	struct pci_dev *pdev = adapter->pdev;
1599
1600	/* Prevent stats update while adapter is being reset */
1601	if (pci_channel_offline(pdev))
1602		return;
1603
1604	if((netdev->flags & IFF_PROMISC) || (netdev->flags & IFF_ALLMULTI) ||
1605	   (netdev->mc_count > IXGB_MAX_NUM_MULTICAST_ADDRESSES)) {
1606		u64 multi = IXGB_READ_REG(&adapter->hw, MPRCL);
1607		u32 bcast_l = IXGB_READ_REG(&adapter->hw, BPRCL);
1608		u32 bcast_h = IXGB_READ_REG(&adapter->hw, BPRCH);
1609		u64 bcast = ((u64)bcast_h << 32) | bcast_l;
1610
1611		multi |= ((u64)IXGB_READ_REG(&adapter->hw, MPRCH) << 32);
1612		/* fix up multicast stats by removing broadcasts */
1613		if(multi >= bcast)
1614			multi -= bcast;
1615
1616		adapter->stats.mprcl += (multi & 0xFFFFFFFF);
1617		adapter->stats.mprch += (multi >> 32);
1618		adapter->stats.bprcl += bcast_l;
1619		adapter->stats.bprch += bcast_h;
1620	} else {
1621		adapter->stats.mprcl += IXGB_READ_REG(&adapter->hw, MPRCL);
1622		adapter->stats.mprch += IXGB_READ_REG(&adapter->hw, MPRCH);
1623		adapter->stats.bprcl += IXGB_READ_REG(&adapter->hw, BPRCL);
1624		adapter->stats.bprch += IXGB_READ_REG(&adapter->hw, BPRCH);
1625	}
1626	adapter->stats.tprl += IXGB_READ_REG(&adapter->hw, TPRL);
1627	adapter->stats.tprh += IXGB_READ_REG(&adapter->hw, TPRH);
1628	adapter->stats.gprcl += IXGB_READ_REG(&adapter->hw, GPRCL);
1629	adapter->stats.gprch += IXGB_READ_REG(&adapter->hw, GPRCH);
1630	adapter->stats.uprcl += IXGB_READ_REG(&adapter->hw, UPRCL);
1631	adapter->stats.uprch += IXGB_READ_REG(&adapter->hw, UPRCH);
1632	adapter->stats.vprcl += IXGB_READ_REG(&adapter->hw, VPRCL);
1633	adapter->stats.vprch += IXGB_READ_REG(&adapter->hw, VPRCH);
1634	adapter->stats.jprcl += IXGB_READ_REG(&adapter->hw, JPRCL);
1635	adapter->stats.jprch += IXGB_READ_REG(&adapter->hw, JPRCH);
1636	adapter->stats.gorcl += IXGB_READ_REG(&adapter->hw, GORCL);
1637	adapter->stats.gorch += IXGB_READ_REG(&adapter->hw, GORCH);
1638	adapter->stats.torl += IXGB_READ_REG(&adapter->hw, TORL);
1639	adapter->stats.torh += IXGB_READ_REG(&adapter->hw, TORH);
1640	adapter->stats.rnbc += IXGB_READ_REG(&adapter->hw, RNBC);
1641	adapter->stats.ruc += IXGB_READ_REG(&adapter->hw, RUC);
1642	adapter->stats.roc += IXGB_READ_REG(&adapter->hw, ROC);
1643	adapter->stats.rlec += IXGB_READ_REG(&adapter->hw, RLEC);
1644	adapter->stats.crcerrs += IXGB_READ_REG(&adapter->hw, CRCERRS);
1645	adapter->stats.icbc += IXGB_READ_REG(&adapter->hw, ICBC);
1646	adapter->stats.ecbc += IXGB_READ_REG(&adapter->hw, ECBC);
1647	adapter->stats.mpc += IXGB_READ_REG(&adapter->hw, MPC);
1648	adapter->stats.tptl += IXGB_READ_REG(&adapter->hw, TPTL);
1649	adapter->stats.tpth += IXGB_READ_REG(&adapter->hw, TPTH);
1650	adapter->stats.gptcl += IXGB_READ_REG(&adapter->hw, GPTCL);
1651	adapter->stats.gptch += IXGB_READ_REG(&adapter->hw, GPTCH);
1652	adapter->stats.bptcl += IXGB_READ_REG(&adapter->hw, BPTCL);
1653	adapter->stats.bptch += IXGB_READ_REG(&adapter->hw, BPTCH);
1654	adapter->stats.mptcl += IXGB_READ_REG(&adapter->hw, MPTCL);
1655	adapter->stats.mptch += IXGB_READ_REG(&adapter->hw, MPTCH);
1656	adapter->stats.uptcl += IXGB_READ_REG(&adapter->hw, UPTCL);
1657	adapter->stats.uptch += IXGB_READ_REG(&adapter->hw, UPTCH);
1658	adapter->stats.vptcl += IXGB_READ_REG(&adapter->hw, VPTCL);
1659	adapter->stats.vptch += IXGB_READ_REG(&adapter->hw, VPTCH);
1660	adapter->stats.jptcl += IXGB_READ_REG(&adapter->hw, JPTCL);
1661	adapter->stats.jptch += IXGB_READ_REG(&adapter->hw, JPTCH);
1662	adapter->stats.gotcl += IXGB_READ_REG(&adapter->hw, GOTCL);
1663	adapter->stats.gotch += IXGB_READ_REG(&adapter->hw, GOTCH);
1664	adapter->stats.totl += IXGB_READ_REG(&adapter->hw, TOTL);
1665	adapter->stats.toth += IXGB_READ_REG(&adapter->hw, TOTH);
1666	adapter->stats.dc += IXGB_READ_REG(&adapter->hw, DC);
1667	adapter->stats.plt64c += IXGB_READ_REG(&adapter->hw, PLT64C);
1668	adapter->stats.tsctc += IXGB_READ_REG(&adapter->hw, TSCTC);
1669	adapter->stats.tsctfc += IXGB_READ_REG(&adapter->hw, TSCTFC);
1670	adapter->stats.ibic += IXGB_READ_REG(&adapter->hw, IBIC);
1671	adapter->stats.rfc += IXGB_READ_REG(&adapter->hw, RFC);
1672	adapter->stats.lfc += IXGB_READ_REG(&adapter->hw, LFC);
1673	adapter->stats.pfrc += IXGB_READ_REG(&adapter->hw, PFRC);
1674	adapter->stats.pftc += IXGB_READ_REG(&adapter->hw, PFTC);
1675	adapter->stats.mcfrc += IXGB_READ_REG(&adapter->hw, MCFRC);
1676	adapter->stats.mcftc += IXGB_READ_REG(&adapter->hw, MCFTC);
1677	adapter->stats.xonrxc += IXGB_READ_REG(&adapter->hw, XONRXC);
1678	adapter->stats.xontxc += IXGB_READ_REG(&adapter->hw, XONTXC);
1679	adapter->stats.xoffrxc += IXGB_READ_REG(&adapter->hw, XOFFRXC);
1680	adapter->stats.xofftxc += IXGB_READ_REG(&adapter->hw, XOFFTXC);
1681	adapter->stats.rjc += IXGB_READ_REG(&adapter->hw, RJC);
1682
1683	/* Fill out the OS statistics structure */
1684
1685	adapter->net_stats.rx_packets = adapter->stats.gprcl;
1686	adapter->net_stats.tx_packets = adapter->stats.gptcl;
1687	adapter->net_stats.rx_bytes = adapter->stats.gorcl;
1688	adapter->net_stats.tx_bytes = adapter->stats.gotcl;
1689	adapter->net_stats.multicast = adapter->stats.mprcl;
1690	adapter->net_stats.collisions = 0;
1691
1692	/* ignore RLEC as it reports errors for padded (<64bytes) frames
1693	 * with a length in the type/len field */
1694	adapter->net_stats.rx_errors =
1695	    /* adapter->stats.rnbc + */ adapter->stats.crcerrs +
1696	    adapter->stats.ruc +
1697	    adapter->stats.roc /*+ adapter->stats.rlec */  +
1698	    adapter->stats.icbc +
1699	    adapter->stats.ecbc + adapter->stats.mpc;
1700
1701	/* see above
1702	 * adapter->net_stats.rx_length_errors = adapter->stats.rlec;
1703	 */
1704
1705	adapter->net_stats.rx_crc_errors = adapter->stats.crcerrs;
1706	adapter->net_stats.rx_fifo_errors = adapter->stats.mpc;
1707	adapter->net_stats.rx_missed_errors = adapter->stats.mpc;
1708	adapter->net_stats.rx_over_errors = adapter->stats.mpc;
1709
1710	adapter->net_stats.tx_errors = 0;
1711	adapter->net_stats.rx_frame_errors = 0;
1712	adapter->net_stats.tx_aborted_errors = 0;
1713	adapter->net_stats.tx_carrier_errors = 0;
1714	adapter->net_stats.tx_fifo_errors = 0;
1715	adapter->net_stats.tx_heartbeat_errors = 0;
1716	adapter->net_stats.tx_window_errors = 0;
1717}
1718
1719#define IXGB_MAX_INTR 10
1720/**
1721 * ixgb_intr - Interrupt Handler
1722 * @irq: interrupt number
1723 * @data: pointer to a network interface device structure
1724 **/
1725
1726static irqreturn_t
1727ixgb_intr(int irq, void *data)
1728{
1729	struct net_device *netdev = data;
1730	struct ixgb_adapter *adapter = netdev_priv(netdev);
1731	struct ixgb_hw *hw = &adapter->hw;
1732	uint32_t icr = IXGB_READ_REG(hw, ICR);
1733#ifndef CONFIG_IXGB_NAPI
1734	unsigned int i;
1735#endif
1736
1737	if(unlikely(!icr))
1738		return IRQ_NONE;  /* Not our interrupt */
1739
1740	if(unlikely(icr & (IXGB_INT_RXSEQ | IXGB_INT_LSC))) {
1741		mod_timer(&adapter->watchdog_timer, jiffies);
1742	}
1743
1744#ifdef CONFIG_IXGB_NAPI
1745	if(netif_rx_schedule_prep(netdev)) {
1746
1747		/* Disable interrupts and register for poll. The flush
1748		  of the posted write is intentionally left out.
1749		*/
1750
1751		atomic_inc(&adapter->irq_sem);
1752		IXGB_WRITE_REG(&adapter->hw, IMC, ~0);
1753		__netif_rx_schedule(netdev);
1754	}
1755#else
1756	/* yes, that is actually a & and it is meant to make sure that
1757	 * every pass through this for loop checks both receive and
1758	 * transmit queues for completed descriptors, intended to
1759	 * avoid starvation issues and assist tx/rx fairness. */
1760	for(i = 0; i < IXGB_MAX_INTR; i++)
1761		if(!ixgb_clean_rx_irq(adapter) &
1762		   !ixgb_clean_tx_irq(adapter))
1763			break;
1764#endif
1765	return IRQ_HANDLED;
1766}
1767
1768#ifdef CONFIG_IXGB_NAPI
1769/**
1770 * ixgb_clean - NAPI Rx polling callback
1771 * @adapter: board private structure
1772 **/
1773
1774static int
1775ixgb_clean(struct net_device *netdev, int *budget)
1776{
1777	struct ixgb_adapter *adapter = netdev_priv(netdev);
1778	int work_to_do = min(*budget, netdev->quota);
1779	int tx_cleaned;
1780	int work_done = 0;
1781
1782	tx_cleaned = ixgb_clean_tx_irq(adapter);
1783	ixgb_clean_rx_irq(adapter, &work_done, work_to_do);
1784
1785	*budget -= work_done;
1786	netdev->quota -= work_done;
1787
1788	/* if no Tx and not enough Rx work done, exit the polling mode */
1789	if((!tx_cleaned && (work_done == 0)) || !netif_running(netdev)) {
1790		netif_rx_complete(netdev);
1791		ixgb_irq_enable(adapter);
1792		return 0;
1793	}
1794
1795	return 1;
1796}
1797#endif
1798
1799/**
1800 * ixgb_clean_tx_irq - Reclaim resources after transmit completes
1801 * @adapter: board private structure
1802 **/
1803
1804static boolean_t
1805ixgb_clean_tx_irq(struct ixgb_adapter *adapter)
1806{
1807	struct ixgb_desc_ring *tx_ring = &adapter->tx_ring;
1808	struct net_device *netdev = adapter->netdev;
1809	struct ixgb_tx_desc *tx_desc, *eop_desc;
1810	struct ixgb_buffer *buffer_info;
1811	unsigned int i, eop;
1812	boolean_t cleaned = FALSE;
1813
1814	i = tx_ring->next_to_clean;
1815	eop = tx_ring->buffer_info[i].next_to_watch;
1816	eop_desc = IXGB_TX_DESC(*tx_ring, eop);
1817
1818	while(eop_desc->status & IXGB_TX_DESC_STATUS_DD) {
1819
1820		for(cleaned = FALSE; !cleaned; ) {
1821			tx_desc = IXGB_TX_DESC(*tx_ring, i);
1822			buffer_info = &tx_ring->buffer_info[i];
1823
1824			if (tx_desc->popts
1825			    & (IXGB_TX_DESC_POPTS_TXSM |
1826			       IXGB_TX_DESC_POPTS_IXSM))
1827				adapter->hw_csum_tx_good++;
1828
1829			ixgb_unmap_and_free_tx_resource(adapter, buffer_info);
1830
1831			*(uint32_t *)&(tx_desc->status) = 0;
1832
1833			cleaned = (i == eop);
1834			if(++i == tx_ring->count) i = 0;
1835		}
1836
1837		eop = tx_ring->buffer_info[i].next_to_watch;
1838		eop_desc = IXGB_TX_DESC(*tx_ring, eop);
1839	}
1840
1841	tx_ring->next_to_clean = i;
1842
1843	if (unlikely(netif_queue_stopped(netdev))) {
1844		spin_lock(&adapter->tx_lock);
1845		if (netif_queue_stopped(netdev) && netif_carrier_ok(netdev) &&
1846		    (IXGB_DESC_UNUSED(tx_ring) >= DESC_NEEDED))
1847			netif_wake_queue(netdev);
1848		spin_unlock(&adapter->tx_lock);
1849	}
1850
1851	if(adapter->detect_tx_hung) {
1852		/* detect a transmit hang in hardware, this serializes the
1853		 * check with the clearing of time_stamp and movement of i */
1854		adapter->detect_tx_hung = FALSE;
1855		if (tx_ring->buffer_info[eop].dma &&
1856		   time_after(jiffies, tx_ring->buffer_info[eop].time_stamp + HZ)
1857		   && !(IXGB_READ_REG(&adapter->hw, STATUS) &
1858		        IXGB_STATUS_TXOFF)) {
1859			/* detected Tx unit hang */
1860			DPRINTK(DRV, ERR, "Detected Tx Unit Hang\n"
1861					"  TDH                  <%x>\n"
1862					"  TDT                  <%x>\n"
1863					"  next_to_use          <%x>\n"
1864					"  next_to_clean        <%x>\n"
1865					"buffer_info[next_to_clean]\n"
1866					"  time_stamp           <%lx>\n"
1867					"  next_to_watch        <%x>\n"
1868					"  jiffies              <%lx>\n"
1869					"  next_to_watch.status <%x>\n",
1870				IXGB_READ_REG(&adapter->hw, TDH),
1871				IXGB_READ_REG(&adapter->hw, TDT),
1872				tx_ring->next_to_use,
1873				tx_ring->next_to_clean,
1874				tx_ring->buffer_info[eop].time_stamp,
1875				eop,
1876				jiffies,
1877				eop_desc->status);
1878			netif_stop_queue(netdev);
1879		}
1880	}
1881
1882	return cleaned;
1883}
1884
1885/**
1886 * ixgb_rx_checksum - Receive Checksum Offload for 82597.
1887 * @adapter: board private structure
1888 * @rx_desc: receive descriptor
1889 * @sk_buff: socket buffer with received data
1890 **/
1891
1892static void
1893ixgb_rx_checksum(struct ixgb_adapter *adapter,
1894		 struct ixgb_rx_desc *rx_desc,
1895		 struct sk_buff *skb)
1896{
1897	/* Ignore Checksum bit is set OR
1898	 * TCP Checksum has not been calculated
1899	 */
1900	if((rx_desc->status & IXGB_RX_DESC_STATUS_IXSM) ||
1901	   (!(rx_desc->status & IXGB_RX_DESC_STATUS_TCPCS))) {
1902		skb->ip_summed = CHECKSUM_NONE;
1903		return;
1904	}
1905
1906	/* At this point we know the hardware did the TCP checksum */
1907	/* now look at the TCP checksum error bit */
1908	if(rx_desc->errors & IXGB_RX_DESC_ERRORS_TCPE) {
1909		/* let the stack verify checksum errors */
1910		skb->ip_summed = CHECKSUM_NONE;
1911		adapter->hw_csum_rx_error++;
1912	} else {
1913		/* TCP checksum is good */
1914		skb->ip_summed = CHECKSUM_UNNECESSARY;
1915		adapter->hw_csum_rx_good++;
1916	}
1917}
1918
1919/**
1920 * ixgb_clean_rx_irq - Send received data up the network stack,
1921 * @adapter: board private structure
1922 **/
1923
1924static boolean_t
1925#ifdef CONFIG_IXGB_NAPI
1926ixgb_clean_rx_irq(struct ixgb_adapter *adapter, int *work_done, int work_to_do)
1927#else
1928ixgb_clean_rx_irq(struct ixgb_adapter *adapter)
1929#endif
1930{
1931	struct ixgb_desc_ring *rx_ring = &adapter->rx_ring;
1932	struct net_device *netdev = adapter->netdev;
1933	struct pci_dev *pdev = adapter->pdev;
1934	struct ixgb_rx_desc *rx_desc, *next_rxd;
1935	struct ixgb_buffer *buffer_info, *next_buffer, *next2_buffer;
1936	uint32_t length;
1937	unsigned int i, j;
1938	boolean_t cleaned = FALSE;
1939
1940	i = rx_ring->next_to_clean;
1941	rx_desc = IXGB_RX_DESC(*rx_ring, i);
1942	buffer_info = &rx_ring->buffer_info[i];
1943
1944	while(rx_desc->status & IXGB_RX_DESC_STATUS_DD) {
1945		struct sk_buff *skb, *next_skb;
1946		u8 status;
1947
1948#ifdef CONFIG_IXGB_NAPI
1949		if(*work_done >= work_to_do)
1950			break;
1951
1952		(*work_done)++;
1953#endif
1954		status = rx_desc->status;
1955		skb = buffer_info->skb;
1956		buffer_info->skb = NULL;
1957
1958		prefetch(skb->data);
1959
1960		if(++i == rx_ring->count) i = 0;
1961		next_rxd = IXGB_RX_DESC(*rx_ring, i);
1962		prefetch(next_rxd);
1963
1964		if((j = i + 1) == rx_ring->count) j = 0;
1965		next2_buffer = &rx_ring->buffer_info[j];
1966		prefetch(next2_buffer);
1967
1968		next_buffer = &rx_ring->buffer_info[i];
1969		next_skb = next_buffer->skb;
1970		prefetch(next_skb);
1971
1972		cleaned = TRUE;
1973
1974		pci_unmap_single(pdev,
1975				 buffer_info->dma,
1976				 buffer_info->length,
1977				 PCI_DMA_FROMDEVICE);
1978
1979		length = le16_to_cpu(rx_desc->length);
1980
1981		if(unlikely(!(status & IXGB_RX_DESC_STATUS_EOP))) {
1982
1983			/* All receives must fit into a single buffer */
1984
1985			IXGB_DBG("Receive packet consumed multiple buffers "
1986					 "length<%x>\n", length);
1987
1988			dev_kfree_skb_irq(skb);
1989			goto rxdesc_done;
1990		}
1991
1992		if (unlikely(rx_desc->errors
1993			     & (IXGB_RX_DESC_ERRORS_CE | IXGB_RX_DESC_ERRORS_SE
1994				| IXGB_RX_DESC_ERRORS_P |
1995				IXGB_RX_DESC_ERRORS_RXE))) {
1996
1997			dev_kfree_skb_irq(skb);
1998			goto rxdesc_done;
1999		}
2000
2001		/* code added for copybreak, this should improve
2002		 * performance for small packets with large amounts
2003		 * of reassembly being done in the stack */
2004#define IXGB_CB_LENGTH 256
2005		if (length < IXGB_CB_LENGTH) {
2006			struct sk_buff *new_skb =
2007			    netdev_alloc_skb(netdev, length + NET_IP_ALIGN);
2008			if (new_skb) {
2009				skb_reserve(new_skb, NET_IP_ALIGN);
2010				skb_copy_to_linear_data_offset(new_skb,
2011							       -NET_IP_ALIGN,
2012							       (skb->data -
2013							        NET_IP_ALIGN),
2014							       (length +
2015							        NET_IP_ALIGN));
2016				/* save the skb in buffer_info as good */
2017				buffer_info->skb = skb;
2018				skb = new_skb;
2019			}
2020		}
2021		/* end copybreak code */
2022
2023		/* Good Receive */
2024		skb_put(skb, length);
2025
2026		/* Receive Checksum Offload */
2027		ixgb_rx_checksum(adapter, rx_desc, skb);
2028
2029		skb->protocol = eth_type_trans(skb, netdev);
2030#ifdef CONFIG_IXGB_NAPI
2031		if(adapter->vlgrp && (status & IXGB_RX_DESC_STATUS_VP)) {
2032			vlan_hwaccel_receive_skb(skb, adapter->vlgrp,
2033				le16_to_cpu(rx_desc->special) &
2034					IXGB_RX_DESC_SPECIAL_VLAN_MASK);
2035		} else {
2036			netif_receive_skb(skb);
2037		}
2038#else /* CONFIG_IXGB_NAPI */
2039		if(adapter->vlgrp && (status & IXGB_RX_DESC_STATUS_VP)) {
2040			vlan_hwaccel_rx(skb, adapter->vlgrp,
2041				le16_to_cpu(rx_desc->special) &
2042					IXGB_RX_DESC_SPECIAL_VLAN_MASK);
2043		} else {
2044			netif_rx(skb);
2045		}
2046#endif /* CONFIG_IXGB_NAPI */
2047		netdev->last_rx = jiffies;
2048
2049rxdesc_done:
2050		/* clean up descriptor, might be written over by hw */
2051		rx_desc->status = 0;
2052
2053		/* use prefetched values */
2054		rx_desc = next_rxd;
2055		buffer_info = next_buffer;
2056	}
2057
2058	rx_ring->next_to_clean = i;
2059
2060	ixgb_alloc_rx_buffers(adapter);
2061
2062	return cleaned;
2063}
2064
2065/**
2066 * ixgb_alloc_rx_buffers - Replace used receive buffers
2067 * @adapter: address of board private structure
2068 **/
2069
2070static void
2071ixgb_alloc_rx_buffers(struct ixgb_adapter *adapter)
2072{
2073	struct ixgb_desc_ring *rx_ring = &adapter->rx_ring;
2074	struct net_device *netdev = adapter->netdev;
2075	struct pci_dev *pdev = adapter->pdev;
2076	struct ixgb_rx_desc *rx_desc;
2077	struct ixgb_buffer *buffer_info;
2078	struct sk_buff *skb;
2079	unsigned int i;
2080	int num_group_tail_writes;
2081	long cleancount;
2082
2083	i = rx_ring->next_to_use;
2084	buffer_info = &rx_ring->buffer_info[i];
2085	cleancount = IXGB_DESC_UNUSED(rx_ring);
2086
2087	num_group_tail_writes = IXGB_RX_BUFFER_WRITE;
2088
2089	/* leave three descriptors unused */
2090	while(--cleancount > 2) {
2091		/* recycle! its good for you */
2092		skb = buffer_info->skb;
2093		if (skb) {
2094			skb_trim(skb, 0);
2095			goto map_skb;
2096		}
2097
2098		skb = netdev_alloc_skb(netdev, adapter->rx_buffer_len
2099			               + NET_IP_ALIGN);
2100		if (unlikely(!skb)) {
2101			/* Better luck next round */
2102			adapter->alloc_rx_buff_failed++;
2103			break;
2104		}
2105
2106		/* Make buffer alignment 2 beyond a 16 byte boundary
2107		 * this will result in a 16 byte aligned IP header after
2108		 * the 14 byte MAC header is removed
2109		 */
2110		skb_reserve(skb, NET_IP_ALIGN);
2111
2112		buffer_info->skb = skb;
2113		buffer_info->length = adapter->rx_buffer_len;
2114map_skb:
2115		buffer_info->dma = pci_map_single(pdev,
2116		                                  skb->data,
2117		                                  adapter->rx_buffer_len,
2118		                                  PCI_DMA_FROMDEVICE);
2119
2120		rx_desc = IXGB_RX_DESC(*rx_ring, i);
2121		rx_desc->buff_addr = cpu_to_le64(buffer_info->dma);
2122		rx_desc->status = 0;
2123
2124
2125		if(++i == rx_ring->count) i = 0;
2126		buffer_info = &rx_ring->buffer_info[i];
2127	}
2128
2129	if (likely(rx_ring->next_to_use != i)) {
2130		rx_ring->next_to_use = i;
2131		if (unlikely(i-- == 0))
2132			i = (rx_ring->count - 1);
2133
2134		/* Force memory writes to complete before letting h/w
2135		 * know there are new descriptors to fetch.  (Only
2136		 * applicable for weak-ordered memory model archs, such
2137		 * as IA-64). */
2138		wmb();
2139		IXGB_WRITE_REG(&adapter->hw, RDT, i);
2140	}
2141}
2142
2143/**
2144 * ixgb_vlan_rx_register - enables or disables vlan tagging/stripping.
2145 *
2146 * @param netdev network interface device structure
2147 * @param grp indicates to enable or disable tagging/stripping
2148 **/
2149static void
2150ixgb_vlan_rx_register(struct net_device *netdev, struct vlan_group *grp)
2151{
2152	struct ixgb_adapter *adapter = netdev_priv(netdev);
2153	uint32_t ctrl, rctl;
2154
2155	ixgb_irq_disable(adapter);
2156	adapter->vlgrp = grp;
2157
2158	if(grp) {
2159		/* enable VLAN tag insert/strip */
2160		ctrl = IXGB_READ_REG(&adapter->hw, CTRL0);
2161		ctrl |= IXGB_CTRL0_VME;
2162		IXGB_WRITE_REG(&adapter->hw, CTRL0, ctrl);
2163
2164		/* enable VLAN receive filtering */
2165
2166		rctl = IXGB_READ_REG(&adapter->hw, RCTL);
2167		rctl |= IXGB_RCTL_VFE;
2168		rctl &= ~IXGB_RCTL_CFIEN;
2169		IXGB_WRITE_REG(&adapter->hw, RCTL, rctl);
2170	} else {
2171		/* disable VLAN tag insert/strip */
2172
2173		ctrl = IXGB_READ_REG(&adapter->hw, CTRL0);
2174		ctrl &= ~IXGB_CTRL0_VME;
2175		IXGB_WRITE_REG(&adapter->hw, CTRL0, ctrl);
2176
2177		/* disable VLAN filtering */
2178
2179		rctl = IXGB_READ_REG(&adapter->hw, RCTL);
2180		rctl &= ~IXGB_RCTL_VFE;
2181		IXGB_WRITE_REG(&adapter->hw, RCTL, rctl);
2182	}
2183
2184	ixgb_irq_enable(adapter);
2185}
2186
2187static void
2188ixgb_vlan_rx_add_vid(struct net_device *netdev, uint16_t vid)
2189{
2190	struct ixgb_adapter *adapter = netdev_priv(netdev);
2191	uint32_t vfta, index;
2192
2193	/* add VID to filter table */
2194
2195	index = (vid >> 5) & 0x7F;
2196	vfta = IXGB_READ_REG_ARRAY(&adapter->hw, VFTA, index);
2197	vfta |= (1 << (vid & 0x1F));
2198	ixgb_write_vfta(&adapter->hw, index, vfta);
2199}
2200
2201static void
2202ixgb_vlan_rx_kill_vid(struct net_device *netdev, uint16_t vid)
2203{
2204	struct ixgb_adapter *adapter = netdev_priv(netdev);
2205	uint32_t vfta, index;
2206
2207	ixgb_irq_disable(adapter);
2208
2209	vlan_group_set_device(adapter->vlgrp, vid, NULL);
2210
2211	ixgb_irq_enable(adapter);
2212
2213	/* remove VID from filter table*/
2214
2215	index = (vid >> 5) & 0x7F;
2216	vfta = IXGB_READ_REG_ARRAY(&adapter->hw, VFTA, index);
2217	vfta &= ~(1 << (vid & 0x1F));
2218	ixgb_write_vfta(&adapter->hw, index, vfta);
2219}
2220
2221static void
2222ixgb_restore_vlan(struct ixgb_adapter *adapter)
2223{
2224	ixgb_vlan_rx_register(adapter->netdev, adapter->vlgrp);
2225
2226	if(adapter->vlgrp) {
2227		uint16_t vid;
2228		for(vid = 0; vid < VLAN_GROUP_ARRAY_LEN; vid++) {
2229			if(!vlan_group_get_device(adapter->vlgrp, vid))
2230				continue;
2231			ixgb_vlan_rx_add_vid(adapter->netdev, vid);
2232		}
2233	}
2234}
2235
2236#ifdef CONFIG_NET_POLL_CONTROLLER
2237/*
2238 * Polling 'interrupt' - used by things like netconsole to send skbs
2239 * without having to re-enable interrupts. It's not called while
2240 * the interrupt routine is executing.
2241 */
2242
2243static void ixgb_netpoll(struct net_device *dev)
2244{
2245	struct ixgb_adapter *adapter = netdev_priv(dev);
2246
2247	disable_irq(adapter->pdev->irq);
2248	ixgb_intr(adapter->pdev->irq, dev);
2249	enable_irq(adapter->pdev->irq);
2250}
2251#endif
2252
2253/**
2254 * ixgb_io_error_detected() - called when PCI error is detected
2255 * @pdev    pointer to pci device with error
2256 * @state   pci channel state after error
2257 *
2258 * This callback is called by the PCI subsystem whenever
2259 * a PCI bus error is detected.
2260 */
2261static pci_ers_result_t ixgb_io_error_detected (struct pci_dev *pdev,
2262			             enum pci_channel_state state)
2263{
2264	struct net_device *netdev = pci_get_drvdata(pdev);
2265	struct ixgb_adapter *adapter = netdev_priv(netdev);
2266
2267	if(netif_running(netdev))
2268		ixgb_down(adapter, TRUE);
2269
2270	pci_disable_device(pdev);
2271
2272	/* Request a slot reset. */
2273	return PCI_ERS_RESULT_NEED_RESET;
2274}
2275
2276/**
2277 * ixgb_io_slot_reset - called after the pci bus has been reset.
2278 * @pdev    pointer to pci device with error
2279 *
2280 * This callback is called after the PCI buss has been reset.
2281 * Basically, this tries to restart the card from scratch.
2282 * This is a shortened version of the device probe/discovery code,
2283 * it resembles the first-half of the ixgb_probe() routine.
2284 */
2285static pci_ers_result_t ixgb_io_slot_reset (struct pci_dev *pdev)
2286{
2287	struct net_device *netdev = pci_get_drvdata(pdev);
2288	struct ixgb_adapter *adapter = netdev_priv(netdev);
2289
2290	if(pci_enable_device(pdev)) {
2291		DPRINTK(PROBE, ERR, "Cannot re-enable PCI device after reset.\n");
2292		return PCI_ERS_RESULT_DISCONNECT;
2293	}
2294
2295	/* Perform card reset only on one instance of the card */
2296	if (0 != PCI_FUNC (pdev->devfn))
2297		return PCI_ERS_RESULT_RECOVERED;
2298
2299	pci_set_master(pdev);
2300
2301	netif_carrier_off(netdev);
2302	netif_stop_queue(netdev);
2303	ixgb_reset(adapter);
2304
2305	/* Make sure the EEPROM is good */
2306	if(!ixgb_validate_eeprom_checksum(&adapter->hw)) {
2307		DPRINTK(PROBE, ERR, "After reset, the EEPROM checksum is not valid.\n");
2308		return PCI_ERS_RESULT_DISCONNECT;
2309	}
2310	ixgb_get_ee_mac_addr(&adapter->hw, netdev->dev_addr);
2311	memcpy(netdev->perm_addr, netdev->dev_addr, netdev->addr_len);
2312
2313	if(!is_valid_ether_addr(netdev->perm_addr)) {
2314		DPRINTK(PROBE, ERR, "After reset, invalid MAC address.\n");
2315		return PCI_ERS_RESULT_DISCONNECT;
2316	}
2317
2318	return PCI_ERS_RESULT_RECOVERED;
2319}
2320
2321/**
2322 * ixgb_io_resume - called when its OK to resume normal operations
2323 * @pdev    pointer to pci device with error
2324 *
2325 * The error recovery driver tells us that its OK to resume
2326 * normal operation. Implementation resembles the second-half
2327 * of the ixgb_probe() routine.
2328 */
2329static void ixgb_io_resume (struct pci_dev *pdev)
2330{
2331	struct net_device *netdev = pci_get_drvdata(pdev);
2332	struct ixgb_adapter *adapter = netdev_priv(netdev);
2333
2334	pci_set_master(pdev);
2335
2336	if(netif_running(netdev)) {
2337		if(ixgb_up(adapter)) {
2338			printk ("ixgb: can't bring device back up after reset\n");
2339			return;
2340		}
2341	}
2342
2343	netif_device_attach(netdev);
2344	mod_timer(&adapter->watchdog_timer, jiffies);
2345}
2346
2347/* ixgb_main.c */
2348